FYP_Pilot2_dataAnalysis

load data: 209 subjects, with 864 goals are included in the following analysis

goalRating_long_R <- read.csv("./output/goalRating_long_R.csv",stringsAsFactors = F)

indivDiffDf <- read.csv("./output/indivDiffDf.csv",stringsAsFactors = F)

goalDf_sum_wide <- read.csv("./output/goalDf_wide.csv",stringsAsFactors = F)

Data Screening for goal representation assessment

Missing data

Check the number of missing data per variable, and below is the top 5 variables. Missing data is rare for all variables

# check the number of "I'm not sure" responses per varialbe
totalGoal <- nrow(goalRating_long_R)/35

goalRating_long_R %>%
  filter(is.na(rating)) %>%
  tabyl(variable) %>%
  mutate(percent = n/totalGoal) %>%
  arrange(desc(percent)) %>%
  head(5)

##                   variable n     percent
## 1              frequency_R 6 0.006944444
## 2            attainability 5 0.005787037
## 3  attractiveness_progress 4 0.004629630
## 4              basic_needs 4 0.004629630
## 5 attainment_maintenance_R 3 0.003472222

The “I’m not sure” response

“construal_level”,“approach_avoidance” and “attainment_maintenance” question have an option for “I’m not sure” because they ask subjects to categorilize their goals.

around 5% of the goals had “I’m not sure” as the response.

The modification for construal level question seemed to be sucessful!

# check the number of "I'm not sure" responses per varialbe
goalRating_long_R %>%
  filter(rating == 99) %>%
  tabyl(variable) %>%
  mutate(percent = n/totalGoal) %>%
  arrange(desc(percent))

##                   variable  n   percent
## 1          construal_level 41 0.0474537
## 2     approach_avoidance_R 40 0.0462963
## 3 attainment_maintenance_R 40 0.0462963

around 3.3% participants select the “I’m not sure” option for construal level for more than once.

# get the number of total subject
totalSub <- nrow(indivDiffDf)

# check the percentage of participants who selected "I'm not sure" for construal level more than once
goalRating_long_R %>%
  filter(rating == 99 & variable == "construal_level") %>%
  tabyl(id) %>%
  filter(n >1) %>%
  nrow()/totalSub

## [1] 0.03349282

The “not specified” response

temporal_duration, frequency and end_state_specificity question have an option for “not specified” because they ask about features that may not be applicable to all goals.

The end state specificity is not applicable to around 17% of the goals

# check the number of "not specified" responses per varialbe
goalRating_long_R %>%
  filter(rating == 999) %>%
  tabyl(variable) %>%
  mutate(percent = n/totalGoal) %>%
  arrange(desc(percent))

##                  variable   n    percent
## 1 end_state_specificity_R 150 0.17361111
## 2       temporal_duration  62 0.07175926
## 3             frequency_R  61 0.07060185

Transform all special cases to NAs

All “I’m not sure” and “not specified” responses will be treated as missing data.

# transform 99 & 999 to NAs
goalRating_long_R <- goalRating_long_R %>% 
  mutate(rating = replace(rating, rating == 99 | rating == 999, NA))

The number of claimed goals

Descriptive on the number of goals subject claimed to have prior to listing them

describe(goalDf_sum_wide$total_goal)

##    vars   n mean   sd median trimmed  mad min max range skew kurtosis   se
## X1    1 209 4.92 2.61      4    4.57 1.48   1  15    14 1.44     2.02 0.18

breaks = (1:15)
goalDf_sum_wide %>% 
  ggplot(aes(x = total_goal)) + 
  scale_x_continuous(labels=scales::comma(breaks, accuracy = 1), breaks=breaks) + 
  geom_histogram(fill = "orange", 
                 colour = "black",
                 binwidth = 1) + 
  labs(x="Number of claimed goals", y="# of participants") +
  theme_classic(base_size = 18) 

The percentage of subjects who claimed having more than 5 goals: 24.4%

length(goalDf_sum_wide$total_goal[goalDf_sum_wide$total_goal>5])/totalSub

## [1] 0.2440191

Descriptive on the number of goals participants actual listed

describe(goalDf_sum_wide$listNum)

##    vars   n mean   sd median trimmed  mad min max range  skew kurtosis   se
## X1    1 209 4.13 1.01      4    4.27 1.48   1   5     4 -0.81    -0.49 0.07

breaks <- (1:5)
goalDf_sum_wide %>% 
  ggplot(aes(x = listNum)) + 
  scale_x_continuous(labels=scales::comma(breaks, accuracy = 1), breaks=seq(1, 5, by = 1)) + 
  geom_histogram(fill = "orange", 
                 colour = "black",
                 binwidth = 1) + 
  labs(x="Number of listed goals", y="# of participants") +
  theme_classic(base_size = 18) 

number of people who listed 1 goal: 1

length(goalDf_sum_wide$listNum[goalDf_sum_wide$listNum == 1])

## [1] 1

descriptvie on the differences between the number of claimed goals and listed goals

goalDf_sum_wide <-goalDf_sum_wide %>%
  mutate(diffNum = total_goal - listNum)

describe(goalDf_sum_wide$diffNum)

##    vars   n mean   sd median trimmed mad min max range skew kurtosis   se
## X1    1 209 0.79 2.11      0    0.45   0  -2  10    12 1.96      4.1 0.15

breaks <- (-2:10)
goalDf_sum_wide %>% 
  ggplot(aes(x = diffNum)) + 
  scale_x_continuous(labels=scales::comma(breaks, accuracy = 1), breaks=breaks) + 
  geom_histogram(fill = "orange", 
                 colour = "black",
                 binwidth = 1) + 
  labs(x="Number of claimed goals - listed goals", y="# of participants") +
  theme_classic(base_size = 18) 

percentage of people who listed more goals than they claimed: 11.5%

length(goalDf_sum_wide$diffNum[goalDf_sum_wide$diffNum <0])/totalSub *100

## [1] 11.48325

percentage of people who listed less goals than they claimed: 26.8%

length(goalDf_sum_wide$diffNum[goalDf_sum_wide$diffNum >0])/totalSub *100

## [1] 26.79426

Even if both the median and the mad of the difference is 0, around 37% of the participants either had to pick 5 out of all of their goals or came up some goals on the spot. Need to be aware of the priming or the order effect.

Goal Representation Ratings

Descriptive stats

# descriptive stats for each variable 
goalRating_long_R %>%
  dplyr::select(variable, rating) %>%
  group_by(variable) %>%
  summarize(mean = mean(rating, na.rm = TRUE),
            sd = sd(rating, na.rm = TRUE), 
            n = n(),
            min = min(rating, na.rm = TRUE),
            max = max(rating, na.rm = TRUE),
            skew = skew(rating, na.rm = T), 
            kurtosi = kurtosi(rating, na.rm = T)
            ) %>%
arrange(skew)

## # A tibble: 35 x 8
##    variable                    mean    sd     n   min   max   skew kurtosi
##    <chr>                      <dbl> <dbl> <int> <int> <int>  <dbl>   <dbl>
##  1 approach_avoidance_R        6.16 1.50    864     1     7 -1.95    3.09 
##  2 control                     6.28 1.05    864     1     7 -1.58    2.29 
##  3 identified_motivation       5.97 1.24    864     1     7 -1.33    1.70 
##  4 ideal_motivation            5.79 1.46    864     1     7 -1.17    0.739
##  5 importance                  6.06 1.16    864     1     7 -1.14    0.675
##  6 social_desirability         5.84 1.31    864     1     7 -1.12    0.858
##  7 attractiveness_achievement  6.12 0.995   864     2     7 -1.10    0.808
##  8 instrumentality             5.57 1.56    864     1     7 -1.08    0.585
##  9 commonality                 5.47 1.61    864     1     7 -1.00    0.311
## 10 initial_time_R              6.31 1.83    864     1     8 -0.982   0.260
## # … with 25 more rows

# order based on their skewness 
#kable(varDf[order(varDf$skew),])

The approach_avoidance has very little variance (most people rated their goals as definately approach goals). After changing the anchors for both the attractiveness variables, their distributions are similar to other variables such as important and ideal_motivation. This time “control” is the most positively skewed variable other than approach_avoidance.

Should we change the approach_avoidance to ordinal?

# histograme for each dimension
goalRating_long_R %>%
  ggplot(aes(x = rating)) +
    geom_histogram(fill   = "orange",
                   colour = "black",
                   alpha  = .6) +
    facet_wrap(~variable, nrow = 7)

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

correlational matrix across all variables

“pairwise.complete.obs” is used for generating correlational matrix.The correlations make sense

# transform the long format to short format
goalDf_wide <- goalRating_long_R %>% spread (variable, rating)

# generate a correlational matrix
corrM_all <- goalDf_wide %>% 
  dplyr :: select(affordance:visibility) %>% 
  cor(use = "pairwise.complete.obs")

# visualization
corrplot(corrM_all, method = "circle",number.cex = .7, order = "AOE", addCoef.col = "black",type = "upper",col= colorRampPalette(c("midnightblue","white", "orange"))(200))

Variance Partition

Only the 30 variables for goal representation are included. Only around 7% of the variance is on the between subject level.

# subset the long format dataset for only the 30 goal representation variable
goal_striving <- c("initial_time_R", "advancement", "urgency", "effort", "commitment")
goalDf_R_long <- goalRating_long_R[!goalRating_long_R$variable %in% goal_striving,]

# generate a multilevel model with subject as the random intercept
mlm <-lmer(rating ~ variable + (1|id), data = goalDf_R_long)

# calculate the variance partition coefficient and transform to ICC
VarCorr(mlm) %>%
  as_tibble() %>%
  mutate(icc=vcov/sum(vcov)) %>%
  dplyr :: select(grp, icc)

## # A tibble: 2 x 2
##   grp         icc
##   <chr>     <dbl>
## 1 id       0.0694
## 2 Residual 0.931

Raw <- VarCorr(mlm) %>%
  as_tibble() %>%
  mutate(Raw=vcov/sum(vcov)) %>%
  dplyr :: select(Raw)

Exploritory Factor Analysis

Data transformation

27 varialbes are included. Ordinal variables are not included: “temporal_duration” & “end_state_specificity” and “frequency”

# Exclude the 5 varialbes related to goal striving progress
goalDf_R_wide <- goalDf_wide[,!names(goalDf_wide) %in% goal_striving]

# Exclude ordinal variables: temporal_duration & end_state_specificity and frequency and other columns with irrelevent data
goal_ordinal <- c("temporal_duration", "end_state_specificity_R", "frequency_R")
goalDf_EFA <- goalDf_R_wide[,!names(goalDf_R_wide) %in% goal_ordinal]
goalDf_EFA <- subset(goalDf_EFA, select = affordance : visibility)

# Generate a correlational matrix 
corrM_raw <- cor(goalDf_EFA, use = "pairwise")

Evaluate the number of factors to extract

Both the Very Simple Structure evaluation and parallel analysis recommend 5 factors. There are 7 factors have an eigen value > 1. Therefore, models with 5, 6, and 7 factors will be explored

# use Very Simple Structure criterion
vss(corrM_raw, n = 10, rotate = "promax", diagonal = FALSE, fm = "minres", 
n.obs=844,plot=TRUE,title="Very Simple Structure",use="pairwise",cor="cor")

## Loading required namespace: GPArotation

## 
## Very Simple Structure
## Call: vss(x = corrM_raw, n = 10, rotate = "promax", diagonal = FALSE, 
##     fm = "minres", n.obs = 844, plot = TRUE, title = "Very Simple Structure", 
##     use = "pairwise", cor = "cor")
## VSS complexity 1 achieves a maximimum of 0.59  with  1  factors
## VSS complexity 2 achieves a maximimum of 0.68  with  5  factors
## 
## The Velicer MAP achieves a minimum of 0.01  with  3  factors 
## BIC achieves a minimum of  -686.51  with  7  factors
## Sample Size adjusted BIC achieves a minimum of  -206.07  with  10  factors
## 
## Statistics by number of factors 
##    vss1 vss2   map dof chisq     prob sqresid  fit RMSEA  BIC  SABIC complex
## 1  0.59 0.00 0.017 324  3375  0.0e+00      25 0.59 0.106 1192 2220.6     1.0
## 2  0.48 0.60 0.014 298  2212 7.7e-289      25 0.60 0.087  204 1149.9     1.3
## 3  0.56 0.67 0.013 273  1637 8.0e-193      20 0.68 0.077 -203  664.0     1.4
## 4  0.55 0.64 0.014 249  1265 1.9e-135      18 0.71 0.070 -413  377.4     1.7
## 5  0.55 0.68 0.014 226   933  1.6e-86      16 0.74 0.061 -590  127.4     1.8
## 6  0.46 0.62 0.017 204   736  3.3e-61      17 0.72 0.056 -639    9.1     1.9
## 7  0.42 0.51 0.020 183   547  7.1e-38      23 0.63 0.049 -687 -105.4     1.8
## 8  0.42 0.51 0.022 163   415  5.7e-24      24 0.61 0.043 -683 -165.3     1.9
## 9  0.43 0.50 0.024 144   322  1.1e-15      23 0.62 0.038 -648 -190.5     1.8
## 10 0.42 0.48 0.028 126   243  2.0e-09      24 0.60 0.033 -606 -206.1     1.9
##    eChisq  SRMR eCRMS eBIC
## 1    5951 0.100 0.104 3768
## 2    3334 0.075 0.081 1326
## 3    1925 0.057 0.065   85
## 4    1155 0.044 0.052 -523
## 5     712 0.035 0.043 -811
## 6     538 0.030 0.040 -836
## 7     386 0.026 0.035 -847
## 8     275 0.022 0.032 -824
## 9     185 0.018 0.028 -785
## 10    120 0.014 0.024 -729

# Use the Scree plot to identify the number of factors have Eigenvalues >1 and the output from the Parallel analysis

ev <- eigen(corrM_raw)
ap <- parallel(subject=nrow(goalDf_EFA),var=ncol(goalDf_EFA),
  rep=100,cent=.05)
nS <- nScree(x=ev$values, aparallel=ap$eigen$qevpea)
plotnScree(nS)

Extract factors

Extract number of factors based on the suggestions above. Because we expect factors to be correlated with each other, we use “promax” rotation.

# extract 5 factors
fa_raw_5 <-fa(r=corrM_raw, nfactors=5,n.obs = 854, rotate="promax", SMC=FALSE, fm="minres")
#fa.sort(fa_raw_5)

# extract 6 factors
fa_raw_6 <-fa(r=corrM_raw, nfactors=6,n.obs = 854, rotate="promax", SMC=FALSE, fm="minres")

# extract 7 factors
fa_raw_7 <-fa(r=corrM_raw, nfactors=7,n.obs = 854, rotate="promax", SMC=FALSE, fm="minres")
#fa.sort(fa_raw_7)

Compare loadings for each model

5 factors

fa.diagram(fa_raw_5)

6 factors

fa.diagram(fa_raw_6)

7 factors

fa.diagram(fa_raw_7)

Compare model fit & complexity

# generate a dataframe 
fa_fitDf <- data.frame(factors = c(5,6,7),
                        chi = c(fa_raw_5$chi,fa_raw_6$chi,fa_raw_7$chi),
                        BIC = c(fa_raw_5$BIC,fa_raw_6$BIC,fa_raw_7$BIC),
                        fit = c(fa_raw_5$fit,fa_raw_6$fit,fa_raw_7$fit),
                        RMSEA = c(fa_raw_5$RMSEA[1],fa_raw_6$RMSEA[1],fa_raw_7$RMSEA[1]),
                       cumVar = c(max(fa_raw_5$Vaccounted[3,]), max(fa_raw_6$Vaccounted[3,]),max(fa_raw_7$Vaccounted[3,])),
                        complexity = c(mean(fa_raw_5$complexity),mean(fa_raw_6$complexity),mean(fa_raw_7$complexity)))

fa_fitDf

##   factors      chi       BIC       fit      RMSEA    cumVar complexity
## 1       5 720.2173 -581.7295 0.8180758 0.06097121 0.4006152   1.755238
## 2       6 544.6918 -632.2759 0.8314411 0.05569833 0.4216004   1.909975
## 3       7 390.9319 -682.0644 0.8450613 0.04865439 0.4464410   1.794470

The most parsimonious model is pretty interpretable. However, more than 1/3 of the variables are clustered on the first factor. This may not be ideal when we explore between subject level variance. The model with 7 factors seems to load (less cross loading) better and explain a little more variance. However, the interfactor correlation may be too high. In order to expand the exploration into the factor variance, I’ll go with 7 factors for now. The additional 2 factors are attractiveness and visibility

Another issue to consider is the mix of unipolar and bipolar. Among the 27 numeric variables, we have 10 bipolar variables: are social desirability, clarity, controllability, all measures about motivations, commonality. They load on different factors on both models, so it may not be a huge issue.

Final model: 7 Factors

Compared to the 7 factor model we saw on April 30th with 854 goals, after adding 10 more goals, “external importance” and “control” loaded with the last factor. “External importance” is loaded pretty evenlly across importance, ought and visibility (0.34, 0.35, 0.37). Control is positively loaded with measuability (0.28) and negatively loaded with visibility (-0.36). The last factor is difficult to interpret in the current model.

# organize loadings
loadings <- fa.sort(fa_raw_7)$loadings
loadings <- as.data.frame(unclass(loadings))
colnames(loadings) <-  c("importance", "ought", "measuarability", "attractiveness", "commonality", "attainability", "external factor")
loadings$Variables <- rownames(loadings)
loadings.m <- loadings %>% gather(-Variables, key = "Factor", value = "Loading")
colOrder <- c("importance", "ought", "measuarability", "attractiveness", "commonality", "attainability", "external factor")
rowOrder <- rev(rownames(loadings))
loadings.m<- arrange(mutate(loadings.m,Variables=factor(Variables,leve=rowOrder)),Variables)
loadings.m<- arrange(mutate(loadings.m,Factor=factor(Factor,leve=colOrder)),Factor)

# Visualization
ggplot(loadings.m, aes(Variables, abs(Loading), fill=Loading)) + 
  facet_wrap(~ Factor, nrow=1) + #place the factors in separate facets
  geom_bar(stat="identity") + #make the bars
  coord_flip() + #flip the axes so the test names can be horizontal  
  #define the fill color gradient: blue=positive, red=negative
  scale_fill_gradient2(name = "Loading", 
                       high = "orange", mid = "white", low = "midnightblue", 
                       midpoint=0, guide="colourbar") +
  ylab("Loading Strength") + #improve y-axis label
  ggtitle("Loadings for 7 factors") + 
  theme_bw(base_size=10)

Alternative model: 5 Factors

# visualization
loadings <- fa.sort(fa_raw_5)$loadings
loadings <- as.data.frame(unclass(loadings))
colnames(loadings) <- c("importance", "ought", "measurability", "attainability", "commonality")
loadings$Variables <- rownames(loadings)
loadings.m <- loadings %>% gather(-Variables, key = "Factor", value = "Loading")
colOrder <- c("importance", "ought", "measurability", "attainability", "commonality")
rowOrder <- rev(rownames(loadings))
loadings.m<- arrange(mutate(loadings.m,Variables=factor(Variables,leve=rowOrder)),Variables)
loadings.m<- arrange(mutate(loadings.m,Factor=factor(Factor,leve=colOrder)),Factor)

ggplot(loadings.m, aes(Variables, abs(Loading), fill=Loading)) + 
  facet_wrap(~ Factor, nrow=1) + #place the factors in separate facets
  geom_bar(stat="identity") + #make the bars
  coord_flip() + #flip the axes so the test names can be horizontal  
  #define the fill color gradient: blue=positive, red=negative
  scale_fill_gradient2(name = "Loading", 
                       high = "orange", mid = "white", low = "midnightblue", 
                       midpoint=0, guide="colourbar") +
  ylab("Loading Strength") + #improve y-axis label + 
  ggtitle("Loadings for 5 factors") + 
  theme_bw(base_size=10)

interfactor correlation: correlation between attractiveness & importance, commonality & importance and ought & commonality are high

fa_raw_7$Phi %>% 
  as.data.frame() %>% 
  dplyr::rename(importance = MR1, ought = MR2, attractiveness = MR6, measuarability = MR3, commonality = MR7, attainability = MR5, external_factor = MR4) %>%
  round(.,2) %>%
  remove_rownames() %>%
  mutate(factor = colnames(.)) %>%
  select(factor, everything()) %>%
  kable(format = "html", escape = F, caption = "Interfactor Correlation") %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = F,position = "center",fixed_thead = T)

Interfactor Correlation

factor	importance	ought	measuarability	attractiveness	commonality	attainability	external_factor
importance	1.00	0.40	0.41	0.53	0.50	0.16	0.23
ought	0.40	1.00	0.16	-0.05	0.48	0.13	0.16
measuarability	0.41	0.16	1.00	0.12	0.11	0.13	0.05
attractiveness	0.53	-0.05	0.12	1.00	0.37	0.15	0.10
commonality	0.50	0.48	0.11	0.37	1.00	0.05	-0.03
attainability	0.16	0.13	0.13	0.15	0.05	1.00	-0.05
external_factor	0.23	0.16	0.05	0.10	-0.03	-0.05	1.00

Factor scores

Generate factors scores using simple weights (0 & 1)

Generate the factor score based on the mean

factorScoreDf <- goalDf_R_wide %>%
  mutate(difficulty = 8 - difficulty,
         control = 8 - control) %>%
  mutate(Importance = rowMeans(select(.,instrumentality, ideal_motivation, connectedness, meaningfulness, identified_motivation, importance, construal_level),na.rm = T),
         Ought = rowMeans(select(., ought_motivation, external_motivation , introjected_motivation), na.rm = T),
         Attractiveness = rowMeans(select(., attractiveness_achievement, attractiveness_progress, intrinsic_motivation), na.rm = T),
         Measurability = rowMeans(select(., measurability , clarity , specificity , approach_avoidance_R), na.rm = T),
         Commonality = rowMeans(select(., commonality , social_desirability, basic_needs), na.rm = T),
         Attainability = rowMeans(select(., difficulty , attainability , affordance), na.rm = T),
         External_factor = rowMeans(select(., visibility , attainment_maintenance_R, control,external_importance), na.rm = T)) %>%
  select(id, list_goal = listNum, claim_goal = total_goal, goal_order = goal, Importance, Ought, Attractiveness, Measurability, Commonality, Attainability, External_factor)

variation partition on factor scores

# Importance 
mlm <-lmer(Importance ~ 1 + (1|id), data = factorScoreDf)

Importance <- VarCorr(mlm) %>%
  as_tibble() %>%
  mutate(Importance=vcov/sum(vcov)) %>%
  dplyr :: select(Importance)

# Ought 
mlm <-lmer(Ought ~ 1 + (1|id), data = factorScoreDf)

Ought <- VarCorr(mlm) %>%
  as_tibble() %>%
  mutate(Ought=vcov/sum(vcov)) %>%
  dplyr :: select(Ought)

# Attractiveness 
mlm <-lmer(Attractiveness ~ 1 + (1|id), data = factorScoreDf)

Attractiveness <- VarCorr(mlm) %>%
  as_tibble() %>%
  mutate(Attractiveness=vcov/sum(vcov)) %>%
  dplyr :: select(Attractiveness)

# Measurability 
mlm <-lmer(Measurability ~ 1 + (1|id), data = factorScoreDf)

Measurability <- VarCorr(mlm) %>%
  as_tibble() %>%
  mutate(Measurability=vcov/sum(vcov)) %>%
  dplyr :: select(Measurability)

# Commonality 
mlm <-lmer(Commonality ~ 1 + (1|id), data = factorScoreDf)

Commonality <- VarCorr(mlm) %>%
  as_tibble() %>%
  mutate(Commonality=vcov/sum(vcov)) %>%
  dplyr :: select(Commonality)

# Attainability 
mlm <-lmer(Attainability ~ 1 + (1|id), data = factorScoreDf)

Attainability <- VarCorr(mlm) %>%
  as_tibble() %>%
  mutate(Attainability=vcov/sum(vcov)) %>%
  dplyr :: select(Attainability)

# External_factor 
mlm <-lmer(External_factor ~ 1 + (1|id), data = factorScoreDf)

External_factor <- VarCorr(mlm) %>%
  as_tibble() %>%
  mutate(External_factor=vcov/sum(vcov)) %>%
  dplyr :: select(External_factor)

# combine the outputs into one data frame
factorScore_icc <- data.frame("variation" = c("between subject", "within subject"))
factorScore_icc <- bind_cols(factorScore_icc, Importance, Ought,Attractiveness, Measurability,Commonality, Attainability, External_factor, Raw)

# visualization
factorV <- factorScore_icc %>% gather(-variation, key = factor, value = ICC)
ggplot(factorV, aes(fill=variation, y=ICC, x=factor)) + 
  geom_bar(position="stack", stat="identity", alpha = .8) +
  scale_fill_manual(values=c("dark blue", "orange")) + 
  ggtitle("ICC output")+ 
  theme(axis.text.x = element_text(angle = 90, hjust = 1))

The correlations between factor scores and goal progress variables

This correlation matrix shows us that most factors are positively correlated with goal progress measures. goal order has a weak correlation with some factor scores (Measurability, Ought)

# combine goal level data
goalLevelDf <- cbind(select(factorScoreDf, "goal_order" : "Commonality"), goalDf_wide[,c("advancement", "commitment", "effort", "urgency", "initial_time_R")])

# generate a correlational matrix 
corrM_goalLevel <- cor(goalLevelDf,use = "pairwise")
res1 <- cor.mtest(goalLevelDf, conf.level = .95)

# visualization
corrplot(corrM_goalLevel, method = "circle",number.cex = .7, order = "AOE", addCoef.col = "black",type = "upper",col= colorRampPalette(c("midnightblue","white", "orange"))(200),p.mat = res1$p, sig.level = .2, tl.col = c("black", "midnightblue", "darkorange1", "darkorange1", "midnightblue", "darkorange1", "midnightblue","midnightblue","midnightblue","darkorange1","darkorange1"))

subject level aggregated factor scores

# calculate the average score for each factor within each subject
subjectDf_factor <- aggregate(. ~ id, data = factorScoreDf, mean)

# calculate the sd for each factor wihtin each subject
subjectDf_factor_sd <- aggregate(. ~ id, data = factorScoreDf, sd) %>% select(-list_goal, -claim_goal)
subjectDf_factor_sd <- subjectDf_factor %>% select(id, list_goal, claim_goal) %>%
  left_join(subjectDf_factor_sd, by = "id")

Mahalanobis distance

This session is for calculaing pair-wise distance between goals within each subject. Because the dimensions are not orthogonal, we use Mahalanobis distance to estimate the distance across all dimensions.

pair-wise distance

We use the covariance matrix across all goal ratings to calculate pair-wise distance. Subjects who only have 1 goal are excluded from this analysis

# set a function for calculating pairwise distiance
mahalanobisFun <- function(df, cov) { 
  MD <- combn(nrow(df), 2, function(x) mahalanobis(as.matrix(df[x[1],]), center = as.matrix(df[x[2],]), cov = cov))
  return(tryCatch(MD, error=function(e) NULL))
}

Exclude subjects with only one goal

# exclude subjects with only one goal
id_oneGoal <- goalDf_wide$id[goalDf_wide$listNum ==1]
factorScoreDf_clean <- factorScoreDf %>% filter(!id %in% id_oneGoal) %>% select(-list_goal, -claim_goal, -goal_order)

# split the dataset by IDs and then get rid off the ID column
splitDf <- split( factorScoreDf_clean, f = factorScoreDf_clean$id)
#splitDf <- split( factorScoreDf, f = factorScoreDf$id)
splitDf <- lapply(splitDf, function(x) subset(x, select = -id))

# get the covariance matrix on factor scores across all goals
factor_cov <- cov(subset(factorScoreDf_clean, select = -id))
#factor_cov <- cov(subset(factorScoreDf, select = -id))

# apply the distance function to each subject
output <- lapply(splitDf, function(x) mahalanobisFun(x, factor_cov))

Average number of pairs per subject: 7. 15 subjects only have 1 pair.

# extract distance values
distance_M <- unlist(output)

# extract the number of pairs per subject
pairNum <- lapply(output, function(x) length(as.vector(x)))
pairNum <- unlist(pairNum)
mean(pairNum)

## [1] 7.014423

# generate a pairwise data frame
id <- unique(factorScoreDf_clean$id)
id_pair <- unlist(mapply(rep, id, pairNum))

# pairId <- unlist(mapply(seq,1,pairNum))

pairDf_M <- data.frame("subject_id" = id_pair,
                     "distance_M" = distance_M)

descriptive

The distance is negatively skewed.

The pair of goals with the least distance is “To get better grades”, “To exercise more”; The pair of goals with the most distance is “My first goal is to get my college degree on time”, “My fourth goal is to be more present in everyday life and to stop worrying about things I cannot control”.

# descriptive of all pairwise distance
describe(pairDf_M$distance_M)

##    vars    n  mean   sd median trimmed  mad min   max range skew kurtosis  se
## X1    1 1459 11.28 7.65   9.39   10.29 6.35 0.1 53.17 53.07 1.39     2.49 0.2

pairDf_M %>% ggplot(aes(x = distance_M)) +
    geom_histogram(fill   = "orange",
                   colour = "black",
                   alpha  = .6)

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

variance partition

Around 37% of the variance in distance is on between subject level

# generate a multilevel model with subject as the random intercept
mlm <-lmer(distance_M ~ 1 + (1|subject_id), data = pairDf_M)

# calculate the variance partition coefficient and transform to ICC
VarCorr(mlm) %>%
  as_tibble() %>%
  mutate(icc=vcov/sum(vcov)) %>%
  dplyr :: select(grp, icc)

## # A tibble: 2 x 2
##   grp          icc
##   <chr>      <dbl>
## 1 subject_id 0.371
## 2 Residual   0.629

subject-level distance (mean & sd)

# calculate mean distance per subject
distDf_perSub_M <- pairDf_M %>%
  group_by(subject_id) %>%
  mutate(distMean = mean(distance_M),
         distSd = sd(distance_M)) %>%
  dplyr :: select(-distance_M)

distDf_perSub_M <- distDf_perSub_M[!duplicated(distDf_perSub_M$subject_id),]

# descriptive of subject-level mean distance
describe(distDf_perSub_M$distMean)

##    vars   n  mean   sd median trimmed  mad  min   max range skew kurtosis  se
## X1    1 208 11.28 5.84  10.44   10.71 4.78 1.07 35.57  34.5 1.18     2.08 0.4

distDf_perSub_M %>% ggplot(aes(x = distMean)) +
    geom_histogram(fill   = "orange",
                   colour = "black",
                   alpha  = .6)

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

# descriptive of subject-level distance sd
describe(distDf_perSub_M$distSd)

##    vars   n mean   sd median trimmed  mad  min   max range skew kurtosis   se
## X1    1 193 5.35 3.08   4.97       5 2.74 0.69 19.56 18.87 1.42     3.32 0.22

distDf_perSub_M %>% ggplot(aes(x = distSd)) +
    geom_histogram(fill   = "orange",
                   colour = "black",
                   alpha  = .6)

## `stat_bin()` using `bins = 30`. Pick better value with `binwidth`.

subject level correlations

Correlations between distance and average factor scores

Both the average and variance of the distance are negatively correlated with most of the average factor scores, indicating that subjects with goals that are on average more important, common, externally motivated and attractive perceive all their goals in a more similar way.

subjectDf <- left_join(subjectDf_factor, distDf_perSub_M, by = c("id" = "subject_id")) %>%
  select(-goal_order)

# generate a correlational matrix 
corrM_factorMean <- cor(subjectDf[,-1],use = "pairwise")
res1 <- cor.mtest(subjectDf[,-1], conf.level = .95)
p_value <- as.data.frame(t(res1$p[c(10,11),3:9]))

# generate and present results in a table
as.data.frame(t(corrM_factorMean[c(10,11),3:9])) %>%
  round(.,2) %>%
  mutate(factor_mean = row.names(.)) %>%
  select(factor_mean, everything()) %>%
  bind_cols(p_value) %>%
  mutate(
    distMean = cell_spec(distMean, "html", color = ifelse(V1 < 0.05, "darkorange", "black")),
    distSd = cell_spec(distSd, "html", color = ifelse(V2 < 0.05, "darkorange", "black"))
  ) %>%
  select(-V1,-V2) %>%
  kable(format = "html", escape = F) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = F,position = "center")

factor_mean	distMean	distSd
Importance	-0.25	-0.15
Ought	-0.18	-0.19
Attractiveness	-0.21	-0.13
Measurability	-0.11	-0.07
Commonality	-0.27	-0.24
Attainability	-0.11	-0.12
External_factor	-0.09	-0.02

Correlations between distance and factor scores variance (sd)

7 factors contribute roughly equally to the average degree and variance of the distance

# combine subject level data frame (individual differences, variance in factor scores, mean and sd in distance)
subjectDf_sd <- left_join(subjectDf_factor_sd, distDf_perSub_M, by = c("id" = "subject_id")) %>%
  select(-goal_order)

# generate a correlational matrix 
corrM_factorSd <- cor(subjectDf_sd[,-1],use = "pairwise")
res1 <- cor.mtest(subjectDf_sd[,-1], conf.level = .95)
p_value <- as.data.frame(t(res1$p[c(10,11),3:9]))

# generate and present results in a table
as.data.frame(t(corrM_factorSd[c(10,11),3:9])) %>%
  round(.,2) %>%
  mutate(factor_sd = row.names(.)) %>%
  select(factor_sd, everything()) %>%
  bind_cols(p_value) %>%
  mutate(
    distMean = cell_spec(distMean, "html", color = ifelse(V1 < 0.05, "darkorange", "black")),
    distSd = cell_spec(distSd, "html", color = ifelse(V2 < 0.05, "darkorange", "black"))
  ) %>%
  select(-V1,-V2) %>%
  kable(format = "html", escape = F) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = F,position = "center")

factor_sd	distMean	distSd
Importance	0.54	0.41
Ought	0.41	0.29
Attractiveness	0.43	0.47
Measurability	0.45	0.35
Commonality	0.5	0.39
Attainability	0.49	0.37
External_factor	0.62	0.37

Correlations between distance and individual measures

Neither the average nor the variance of the distance show clear correlations with individual measures. The number of claimed goals is positively correlate with distance variance but the number of listed goals doesn’t have the same relationship. Some subscales from the “Goal orientation” scale and the “Contingencies of Self-Worth Scale” correlate with distance but the relationship is very weak.

subjectDf <- left_join(indivDiffDf, distDf_perSub_M, by = c("id" = "subject_id"))
subjectDf <- subjectDf_factor %>%
  select(id, list_goal, claim_goal) %>%
  left_join(subjectDf, by = "id")

# generate a correlational matrix 
corrM_im <- cor(subjectDf[,-1],use = "pairwise")
res1 <- cor.mtest(subjectDf[,-1], conf.level = .95)
p_value <- as.data.frame(t(res1$p[c(24,25),1:23]))

# generate and present results in a table
as.data.frame(t(corrM_im[c(24,25),1:23])) %>%
  round(.,2) %>%
  mutate(measures = row.names(.)) %>%
  select(measures, everything()) %>%
  bind_cols(p_value) %>%
  mutate(
    distMean = cell_spec(distMean, "html", color = ifelse(V1 < 0.05, "darkorange", "black")),
    distSd = cell_spec(distSd, "html", color = ifelse(V2 < 0.05, "darkorange", "black"))
  ) %>%
  select(-V1,-V2) %>%
  kable(format = "html", escape = F) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = F,position = "center",fixed_thead = T)

measures	distMean	distSd
list_goal	0.01	0.08
claim_goal	0.09	0.15
Extraversion_mean	0.05	0
Agreeableness_mean	-0.06	-0.13
Conscientiousness_mean	0.03	-0.06
Neuroticism_mean	0.04	0.08
OpenMindedness_mean	0.04	0.05
BSCS_mean	-0.01	0
GOS_learning	0.14	0.19
GOS_avoidance	-0.04	-0.12
GOS_prove	-0.1	-0.15
GSE_mean	0.06	0.04
LET_mean	-0.03	-0.09
PS_mean	0.12	0.01
RSE_mean	0	0
SWL_mean	-0.12	-0.09
family_mean	0	-0.03
competetion_mean	0.14	0.12
appearance_mean	0.02	0.04
god_mean	-0.14	-0.14
academic_mean	-0.07	-0.05
virtue_mean	-0.11	-0.01
approval_mean	0.01	0.03

Correlations between average factor scores and individual measures

Attainability positively correlates with agreeableness, conscientiousness, self-efficacy, life engagement, plantfullness, and negatively correlate with neuroticism. Self esteem and subjective well being show similar pattern: positively correlate with attainability and negatively correlate with ought.

subjectDf <- left_join(indivDiffDf, subjectDf_factor, by = "id") %>% select(-goal_order, -id)

# generate a correlational matrix 
corrM_factorMean_im <- cor(subjectDf,use = "pairwise")
res1 <- cor.mtest(subjectDf, conf.level = .95)
p_value <- as.data.frame(t(res1$p[24:30,1:23]))

# generate and present results in a table
as.data.frame(t(corrM_factorMean_im[24:30,1:23])) %>%
  round(.,2) %>%
  mutate(measures = row.names(.)) %>%
  select(measures, everything()) %>%
  bind_cols(p_value) %>%
  mutate(
    Importance = cell_spec(Importance, "html", color = ifelse(V1 < 0.05, "darkorange", "black")),
    Ought = cell_spec(Ought, "html", color = ifelse(V2 < 0.05, "darkorange", "black")),
    Attractiveness = cell_spec(Attractiveness, "html", color = ifelse(V3 < 0.05, "darkorange", "black")),
    Measurability = cell_spec(Measurability, "html", color = ifelse(V4 < 0.05, "darkorange", "black")),
    Commonality = cell_spec(Commonality, "html", color = ifelse(V5 < 0.05, "darkorange", "black")),
    Attainability = cell_spec(Attainability, "html", color = ifelse(V6 < 0.05, "darkorange", "black")),
    External_factor = cell_spec(External_factor, "html", color = ifelse(V7 < 0.05, "darkorange", "black")),
  ) %>%
  select(-contains("V")) %>%
  kable(format = "html", escape = F, caption = "Correlations between the average factor scores and individual measures") %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = F,position = "center",fixed_thead = T)

Correlations between the average factor scores and individual measures

measures	Importance	Ought	Measurability	Commonality	Attainability	External_factor
Extraversion_mean	0.12	-0.05	0.07	0.06	0.12	0.02
Agreeableness_mean	0.12	-0.01	0.04	0.08	0.19	-0.02
Conscientiousness_mean	0.05	-0.11	-0.02	0.01	0.14	0.03
Neuroticism_mean	-0.04	0.1	-0.05	-0.12	-0.3	-0.09
OpenMindedness_mean	0.18	0.08	0	0.08	0.03	0.01
BSCS_mean	0.04	-0.08	0.1	-0.07	0.26	0.05
GOS_learning	0.18	-0.13	0.22	-0.06	0.25	-0.05
GOS_avoidance	-0.03	0.29	-0.07	0.02	-0.17	0.02
GOS_prove	0.23	0.19	0.11	0.13	-0.03	0.07
GSE_mean	0.17	-0.01	0.17	0.16	0.4	0.05
LET_mean	0.2	-0.14	0.24	0.18	0.27	0.01
PS_mean	0.1	-0.03	0.16	0.09	0.19	0.04
RSE_mean	0.06	-0.25	0.16	0.1	0.3	0.01
SWL_mean	0.09	-0.14	0.12	0.1	0.34	-0.04
family_mean	0.18	0.13	0.17	0.04	0.03	0.1
competetion_mean	0.05	0.17	0.09	-0.02	-0.08	-0.05
appearance_mean	-0.06	0.03	0.02	-0.07	-0.09	-0.04
god_mean	0.17	0.07	0.12	0.19	-0.01	0.06
academic_mean	0.17	0.18	0.13	0.04	-0.17	0.03
virtue_mean	0.19	0.05	0.16	0.17	0.06	-0.01
approval_mean	0.03	0.17	-0.01	-0.06	-0.18	0.06
list_goal	0.06	-0.1	-0.14	-0.05	-0.06	-0.07
claim_goal	0.07	-0.1	-0.08	-0.06	0.04	0

Correlations between factor score variance (sd) and individual measures

The nubmer of listed goals is correlated with variance in measurability, commonality and attainability, but not with claimed goals.

subjectDf <- left_join(indivDiffDf, subjectDf_factor_sd, by = "id") %>% select(-goal_order, -id)

# generate a correlational matrix 
corrM_factorSd_im <- cor(subjectDf,use = "pairwise")
res1 <- cor.mtest(subjectDf, conf.level = .95)
p_value <- as.data.frame(t(res1$p[24:30,1:23]))

# generate and present results in a table
as.data.frame(t(corrM_factorSd_im[24:30,1:23])) %>%
  round(.,2) %>%
  mutate(measures = row.names(.)) %>%
  select(measures, everything()) %>%
  bind_cols(p_value) %>%
  mutate(
    Importance = cell_spec(Importance, "html", color = ifelse(V1 < 0.05, "darkorange", "black")),
    Ought = cell_spec(Ought, "html", color = ifelse(V2 < 0.05, "darkorange", "black")),
    Attractiveness = cell_spec(Attractiveness, "html", color = ifelse(V3 < 0.05, "darkorange", "black")),
    Measurability = cell_spec(Measurability, "html", color = ifelse(V4 < 0.05, "darkorange", "black")),
    Commonality = cell_spec(Commonality, "html", color = ifelse(V5 < 0.05, "darkorange", "black")),
    Attainability = cell_spec(Attainability, "html", color = ifelse(V6 < 0.05, "darkorange", "black")),
    External_factor = cell_spec(External_factor, "html", color = ifelse(V7 < 0.05, "darkorange", "black")),
  ) %>%
  select(-contains("V")) %>%
  kable(format = "html", escape = F, caption = "Correlations between the factor score sd and individual measures") %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = F,position = "center",fixed_thead = T)

Correlations between the factor score sd and individual measures

measures	Importance	Ought	Measurability	Commonality	Attainability	External_factor
Extraversion_mean	-0.11	-0.1	-0.03	0.07	0.01	0.05
Agreeableness_mean	-0.02	-0.1	-0.14	0.03	0.01	-0.06
Conscientiousness_mean	-0.01	-0.02	0.01	0.03	-0.1	0.13
Neuroticism_mean	0.02	0	0	0.03	0.09	0
OpenMindedness_mean	-0.04	0.01	0.01	0.14	0.03	0
BSCS_mean	-0.02	-0.07	-0.06	0.04	-0.1	0.05
GOS_learning	0.08	0.03	0.03	0.22	0.06	0.11
GOS_avoidance	-0.07	-0.03	-0.1	-0.07	0	-0.09
GOS_prove	-0.12	-0.03	-0.11	0.02	0.07	-0.02
GSE_mean	-0.01	-0.07	0.03	0.03	-0.02	0.05
LET_mean	0	0.07	-0.16	0.04	0.04	0.01
PS_mean	0.03	-0.04	0	0.07	0.01	0.21
RSE_mean	0	-0.03	-0.05	0.01	-0.01	0.04
SWL_mean	0.01	-0.02	-0.05	-0.01	0.02	-0.08
family_mean	-0.02	-0.08	-0.02	0.1	0.07	-0.03
competetion_mean	0.02	-0.02	0.13	0.04	0.08	0.04
appearance_mean	0.05	-0.08	0.01	0	0.03	0.04
god_mean	-0.1	-0.11	-0.11	-0.02	0.01	-0.16
academic_mean	0.01	-0.13	0.01	0	0.08	-0.09
virtue_mean	0	-0.05	0.02	-0.01	0.05	-0.12
approval_mean	-0.02	0.03	0.05	0	-0.02	-0.03
list_goal	0.12	0.08	0.25	0.15	0.18	0
claim_goal	0.14	0.11	0.17	0.11	0.09	-0.04

after excluding outliers in distMean(beyond 3 times of sd)

Exclude outliers

# extract outlier IDs for distMean
m <- mean(distDf_perSub_M$distMean, na.rm = T)
sd <- sd(distDf_perSub_M$distMean, na.rm = T)

outlier_mean <- distDf_perSub_M$subject_id[distDf_perSub_M$distMean > m + 3*sd]

# extract outlier IDs for distSd
m <- mean(distDf_perSub_M$distSd, na.rm = T)
sd <- sd(distDf_perSub_M$distSd, na.rm = T)

outlier_sd <- distDf_perSub_M$subject_id[distDf_perSub_M$distSd > m + 3*sd]

Correlation between distMean and all other measures

subjectDf <- left_join(subjectDf_factor, distDf_perSub_M, by = c("id" = "subject_id")) %>%
  left_join(indivDiffDf, by = "id") %>%
  filter(! id %in% outlier_mean) %>%
  select(-id,-goal_order)

# generate a correlational matrix 
corrM_distMean <- cor(subjectDf,use = "pairwise")
res1 <- cor.mtest(subjectDf, conf.level = .95)

# extract p value
p <- res1$p[,10]
p_value <- as.data.frame(p)

# generate table
distMean <-corrM_distMean[,"distMean"] 
meanCorrDf <- as.data.frame(distMean) %>%
  round(.,2) %>%
  mutate(measures = row.names(.)) %>%
  select(measures, everything()) %>%
  bind_cols(p_value) %>%
  mutate(
    distMean = cell_spec(distMean, "html", color = ifelse(p < 0.05, "darkorange", "black"))) %>%
  select(-p)

Correlation between distSd and all other measures

subjectDf <- left_join(subjectDf_factor, distDf_perSub_M, by = c("id" = "subject_id")) %>%
  left_join(indivDiffDf, by = "id") %>%
  filter(! id %in% outlier_sd) %>%
  select(-id,-goal_order)

# generate a correlational matrix 
corrM_distSd <- cor(subjectDf,use = "pairwise")
res1 <- cor.mtest(subjectDf, conf.level = .95)

# extract p value
p <- res1$p[,11]
p_value <- as.data.frame(p)

# generate table
distSd <-corrM_distMean[,"distSd"] 
sdCorrDf <- as.data.frame(distSd) %>%
  round(.,2) %>%
  mutate(measures = row.names(.)) %>%
  select(measures, everything()) %>%
  bind_cols(p_value) %>%
  mutate(
    distSd = cell_spec(distSd, "html", color = ifelse(p < 0.05, "darkorange", "black"))) %>%
  select(-p)

Correlations between distance and other subject level measures after excluding the outliers

left_join(meanCorrDf, sdCorrDf, by = "measures") %>%
  kable(format = "html", escape = F) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = F,position = "center",fixed_thead = T)

measures	distMean	distSd
list_goal	0.1	0.09
claim_goal	0.15	0.16
Importance	-0.22	-0.15
Ought	-0.17	-0.18
Attractiveness	-0.22	-0.14
Measurability	-0.13	-0.06
Commonality	-0.3	-0.24
Attainability	-0.11	-0.11
External_factor	-0.08	-0.02
distMean	1	0.82
distSd	0.82	1
Extraversion_mean	-0.01	-0.02
Agreeableness_mean	0	-0.11
Conscientiousness_mean	0	-0.06
Neuroticism_mean	0.03	0.06
OpenMindedness_mean	0.06	0.05
BSCS_mean	0	0.01
GOS_learning	0.12	0.19
GOS_avoidance	-0.08	-0.12
GOS_prove	-0.11	-0.15
GSE_mean	0	0.01
LET_mean	-0.05	-0.08
PS_mean	0.01	-0.03
RSE_mean	-0.04	0
SWL_mean	-0.09	-0.05
family_mean	0.02	-0.05
competetion_mean	0.13	0.11
appearance_mean	0.05	0.04
god_mean	-0.1	-0.13
academic_mean	0	-0.01
virtue_mean	-0.03	0.03
approval_mean	0.07	0.03

explore the effect of goal listing task

Comparing people who were able to listed all their goals vs. those who had to choose 5 or less out of all their goals.

Differences in distance across 3 different groups

Seperate subjects into 3 groups (list = claimed; list < claimed; list > claimed)

# extract ID numbers
id_claimMore <- unique(goalDf_sum_wide$id[goalDf_sum_wide$diffNum > 0])
id_same <- unique(goalDf_sum_wide$id[goalDf_sum_wide$diffNum == 0])
id_listMore <- unique(goalDf_sum_wide$id[goalDf_sum_wide$diffNum < 0])

# assign group
distDf_perSub_M <- distDf_perSub_M %>%
  mutate(group = case_when(
    subject_id %in% id_claimMore ~ "claimMore",
    subject_id %in% id_same ~ "same",
    subject_id %in% id_listMore ~ "listMore"
  ))

Compare the average and variance of distance across 3 groups

distDf_perSub_M %>%
  group_by(group) %>%
  summarise(
    count = n(),
    distMean = mean(distMean, na.rm = TRUE),
    distSd = mean(distSd, na.rm = TRUE)
  ) %>%
  kable(format = "html", escape = F, caption = "Group mean of the average distance and distance variance") %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = F,position = "center",fixed_thead = T)

Group mean of the average distance and distance variance

group	count	distMean	distSd
claimMore	56	12.18542	6.165695
listMore	24	10.76212	5.471476
same	128	10.97886	4.932374

distDf_perSub_M %>%
  ggplot(aes(group, distMean)) +
  geom_violin() +
  stat_summary(fun=mean, geom="point", shape=23, size=2) +
  ggtitle("distMean")

distDf_perSub_M %>%
  ggplot(aes(group, distSd)) +
  geom_violin() +
  stat_summary(fun=mean, geom="point", shape=23, size=2) +
  ggtitle("distSd")

## Warning: Removed 15 rows containing non-finite values (stat_ydensity).

## Warning: Removed 15 rows containing non-finite values (stat_summary).

The average distance doesn’t differenciate across groups

res.aov <- aov(distMean ~ group, data = distDf_perSub_M)
summary(res.aov)

##              Df Sum Sq Mean Sq F value Pr(>F)
## group         2     64   31.98   0.938  0.393
## Residuals   205   6990   34.10

People who were able to list all their goals have more variance in their pair-wise distance

res.aov <- aov(distSd ~ group, data = distDf_perSub_M)
summary(res.aov)

##              Df Sum Sq Mean Sq F value Pr(>F)  
## group         2   56.8  28.416   3.059 0.0492 *
## Residuals   190 1764.9   9.289                 
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 15 observations deleted due to missingness

Explore correlation between distance and individual measures only among people in the same group

Similar to those using all subjects, the subject-level relationship is weak. Plantfulness is positively correlated with average distance.

subjectDf <- left_join(indivDiffDf, distDf_perSub_M, by = c("id" = "subject_id"))
subjectDf <- subjectDf_factor %>%
  select(id, list_goal, claim_goal) %>%
  left_join(subjectDf, by = "id") %>%
  filter(id %in% id_same) %>%
  select(-id,-group)

# generate a correlational matrix
corrM_im <- cor(subjectDf,use = "pairwise")
res1 <- cor.mtest(subjectDf, conf.level = .95)
p_value <- as.data.frame(t(res1$p[c(24,25),1:23]))

# generate and present results in a table
as.data.frame(t(corrM_im[c(24,25),1:23])) %>%
  round(.,2) %>%
  mutate(measures = row.names(.)) %>%
  select(measures, everything()) %>%
  bind_cols(p_value) %>%
  mutate(
    distMean = cell_spec(distMean, "html", color = ifelse(V1 < 0.05, "darkorange", "black")),
    distSd = cell_spec(distSd, "html", color = ifelse(V2 < 0.05, "darkorange", "black"))
  ) %>%
  select(-V1,-V2) %>%
  kable(format = "html", escape = F) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = F,position = "center",fixed_thead = T)

measures	distMean	distSd
list_goal	0.01	0.1
claim_goal	0.01	0.1
Extraversion_mean	0.1	-0.04
Agreeableness_mean	-0.01	-0.07
Conscientiousness_mean	0.09	-0.05
Neuroticism_mean	0.04	0.06
OpenMindedness_mean	0.09	0.09
BSCS_mean	0.1	0.16
GOS_learning	0.19	0.19
GOS_avoidance	-0.02	-0.11
GOS_prove	0.03	-0.03
GSE_mean	0.08	0.06
LET_mean	0.12	0.01
PS_mean	0.25	0.16
RSE_mean	0.06	0.05
SWL_mean	-0.08	-0.01
family_mean	0.01	-0.03
competetion_mean	0.18	0.15
appearance_mean	0.02	0.13
god_mean	-0.1	-0.19
academic_mean	-0.05	0
virtue_mean	-0.12	0.04
approval_mean	-0.01	-0.03

Differences in each factor score across group

Factor scores are not different across groups. The range in the same group extended to the lower end for some factor.

subjectDf_factor <- subjectDf_factor %>%
  mutate(group = case_when(
    id %in% id_claimMore ~ "claimMore",
    id %in% id_same ~ "same",
    id %in% id_listMore ~ "listMore"
  ))

subjectDf_factor %>%
  group_by(group) %>%
  summarise(
    Importance = mean(Importance, na.rm = TRUE),
    Ought = mean(Ought, na.rm = TRUE),
    Attractiveness = mean(Attractiveness, na.rm = TRUE),
    Measurability = mean(Measurability, na.rm = TRUE),
    Commonality = mean(Commonality, na.rm = TRUE),
    Attainability = mean(Attainability, na.rm = TRUE),
    External_factor = mean(External_factor, na.rm = TRUE),
  ) %>%
  kable(format = "html", escape = F) %>%
  kable_styling(bootstrap_options = c("striped", "hover", "condensed"), full_width = F,position = "center",fixed_thead = T)

group	Importance	Ought	Attractiveness	Measurability	Commonality	Attainability	External_factor
claimMore	5.391157	3.888095	5.447619	5.409375	5.387798	5.533234	3.761309
listMore	5.455274	4.328819	5.382639	5.307465	5.409259	5.538426	3.814873
same	5.318688	3.997158	5.334195	5.470962	5.520004	5.510422	3.656223

subjectDf_factor %>%
  gather(Importance: External_factor, key = "factor", value = "factorScore") %>%
  ggplot(aes(group, factorScore)) +
  geom_violin() +
  stat_summary(fun=mean, geom="point", shape=23, size=2) +
  facet_wrap(~factor, nrow = 4)